How FameX works

Short version: FameX searches YouTube with the official YouTube Data API, keeps channels that are active and growing, asks an AI vision model to check whether each one is faceless and cites the evidence, then tracks every published channel daily. Subscriber and view counts are YouTube data. Verdicts, monetization status, revenue and opportunity scores are FameX estimates.

1. Discovery

Every hour FameX runs YouTube searches for a rotating set of several hundred keyword seeds in English and other languages, varying the sort order and video length. Two more sources add channels: niche tags that keep appearing on published channels become new seeds automatically, and channels featured by published channels are checked too.

A found channel is kept as a candidate only if it is active: it must upload regularly and have a minimum number of subscribers. Everything else is discarded before any AI analysis.

2. Faceless verification

For each candidate FameX builds an evidence pack: the avatar, the banner, the ten most recent thumbnails, the About text, and recent titles and descriptions. A vision-capable AI model classifies the channel into one of eight faceless formats (see channel types) or “not faceless”, gives a verdict (confirmed, likely, unclear) with a confidence score, and must cite concrete observations, for example “the same synthetic face in 9 of 10 thumbnails”.

• Confirmed or likely verdicts with at least 70% confidence are published.
• Unclear or low-confidence results go to a human review queue.
• Channels where a real person presents on camera are rejected.

The first 775 channels in the library were imported from avatarchannels.com with attribution; those pages say so and are not indexed until FameX has analysed them itself.

3. Tracking and freshness

Published channels are refreshed several times a day. FameX stores one statistics snapshot per channel per day, which powers the 7-day and 30-day growth figures; growth is never extrapolated. Recent videos are re-fetched every 12 hours. YouTube data shown on FameX is always less than 30 days old, as the YouTube API Services policies require.

4. Monetization estimate

YouTube does not publish which channels are in the YouTube Partner Program, so FameX estimates it against the public thresholds: 1,000 subscribers and 4,000 long-form watch hours in the last 12 months.

• Watch hours = estimated views in the last 12 months × share of views on long-form videos × median video length × a typical retention rate for that length (45% for videos up to 8 minutes, falling to 22% for videos over 40 minutes).
• Views in the last 12 months use the larger of lifetime views per year and views on recent uploads, or tracked 30-day views × 12 once FameX has a month of snapshots.
• Channels marked “made for kids” are shown separately, because they can only run limited ads.

An admin can mark a channel as verified monetized or not monetized; those pages say “verified”.

5. Revenue estimate

Monthly ad revenue = views in the last 30 days ÷ 1,000 × the typical RPM range for the channel's niche. It is shown as a range and only for channels that look monetized (a quarter of the range for made-for-kids channels). Sponsorships, memberships and product sales are not included.

6. Niche opportunity score

Each niche gets a score from 0 to 100 that compares it with the other niches FameX tracks. It is a weighted percentile rank of four signals:

Scores of 65 and above are labelled High, 40–64 Medium. Niches with fewer than 8 tracked channels are flagged as low confidence.

7. Limits

• The library is not a census of YouTube; it is weighted toward the keywords FameX searches.
• AI verdicts can be wrong. Every channel page shows the evidence so you can judge it yourself.
• Revenue and monetization are estimates from public data and can differ from what a creator actually earns.